Download in silico for prediction of resistance

focuS on PErSonalizEd MEdicinE
Targeted therapy in giST: in silico modeling
for prediction of resistance
Marco A. Pierotti, Elena Tamborini, Tiziana Negri, Sabrina Pricl and Silvana Pilotti
abstract | Elucidation of the genetic processes leading to neoplastic transformation has identified cancerpromoting molecular alterations that can be selectively targeted by rationally designed therapeutic agents.
Protein kinases are druggable targets and have been studied intensively. New methodologies—including
crystallography and three-dimensional modeling—have allowed the rational design of potent and selective
kinase inhibitors that have already reached the clinical stage. However, despite the clinical success of
kinase-targeted therapies, most patients that respond eventually relapse as a result of acquired resistance.
Darwinian-type selection of secondary mutations seems to have a major role in this resistance. The
emergence and/or expansion of tumor clones containing new mutations in the target kinase and that are
drug-insensitive have been observed after chronic treatment. The resistance mechanisms to tyrosine kinase
inhibitors, in particular secondary resistant mutations as a consequence of treatment, will be discussed
in detail. In particular, this Review will focus on KIT and PDGFRA mutations, which are involved in the
pathogenesis of gastrointestinal stromal tumors. Harnessing the selection of mutated variants developed
to overcome these resistance mechanisms is an ongoing goal of current research and new strategies to
overcome drug resistance is being envisaged.
Pierotti, M. A. et al. Nat. Rev. Clin. Oncol. 8, 161–170 (2011); doi:10.1038/nrclinonc.2011.3
Introduction
Protein kinases are commonly implicated in cancer.1 the
human genome encodes more than 500 protein kinases,
many of which have been identified as interesting targets
for drug discovery. nonetheless, few kinase inhibitors
have been licensed and, despite early clinical success, drug
resistance has been observed in patients who initially
responded to such therapies. the resumption of tumor
growth as a consequence of acquired drug resistance may
be promoted by different mechanisms, the most common
being the presence of secondary mutations in the target
tyrosine kinase. these genetic alterations and resulting
protein sequence changes often alter the structure of the
kinase catalytic domain, causing it to have a lower affinity for the inhibitor. such changes have been suggested
by preclinical studies as well as by gene sequencing in
patients who developed secondary resistance in the clinical setting.2 the intrinsic properties of these mutations
can affect the structural state of protein kinases, their
affinity for atP and catalytic activity. moreover, these
mutations can exert global dynamic effects on the structure of the kinase (active and inactive form), alter its
binding-site characteristics and change the movements
of amino acid residues. understanding these aspects can
shed new light on drug resistance mechanisms.
in this context, a relatively fast and noninvasive global
approach based on an in silico analysis (that is, using
competing interests
The authors declare no competing interests.
computer-based molecular simulation) represents a
useful tool that, coupled with traditional biochemical
molecular evidence, may help overcome this problem.
In silico analysis, which uses the three-dimensional (3D)
structure of the receptor protein kinase, has multiple uses.
the method can predict structural changes introduced
by mutations, measure the strength of the interactions
between the protein model and the drug, determine
whether the drug can be effective or not by calculating
the binding affinity (or binding free energy), and evaluate
to what extent decreasing efficacy could still be counteracted by increasing the drug dose or by using a different
compound (supplementary Box 1 online).
in this review, we discuss Kit and PDGFra receptor
tyrosine kinases (rtKs) in the context of gastrointestinal
stromal tumors (Gists, Box 1). we focus on the major
challenge in kinase drug discovery—that is, the emergence of resistance—and discuss techniques to predict
and help prevent this adverse clinical event.
RTK structure and function
rtKs consist of an extracellular domain, a transmembrane
domain and an intracellular region. Physiologically, rtK
activation resulting in phosphorylation in intracellular
tyrosine residues that is achieved by binding cognate
ligands, triggers a cascade of biological reactions leading
to the on/off switch of the genes involved in cell growth,
differentiation and survival. Deregulation of rtK function
may be caused by gene mutation, amplification, protein
nature reviews | clinical oncology
Scientific Directorate
(M. a. Pierotti),
Laboratory of
Experimental Molecular
Pathology
(E. Tamborini, T. negri,
S. Pilotti), Fondazione
IRCCS Istituto
Nazionale dei TumoriMilano, via Venezian 1,
20133 Milano, Italy.
Molecular Simulation
Engineering Laboratory,
DI3, University of
Trieste, Piazzale
Europa 1, 34127
Trieste, Italy (S. Pricl).
Correspondence to:
S. Pricl
sabrina.pricl@
dicamp.units.it
volume 8 | marCH 2011 | 161
© 2011 Macmillan Publishers Limited. All rights reserved
rEViEWS
Key points
■ The pathogenetic role of constitutively activated receptor tyrosine kinases
(RTKs) and drugs that specifically target this alteration in cancer has provided a
new therapeutic opportunity
■ Despite encouraging early therapeutic results, the development of resistance
can occur after a variable period of chronic treatment
■ The emergence of secondary mutations that affect the tyrosine kinase domain
of RTKs reduce the drug binding affinity to the enzymatic pocket of the receptor;
this resistance can be overcome by the development of drugs that bind
efficiently the new mutated RTK forms
■ A new in silico approach is consistent with both biochemical and molecular data
and patient clinical outcome and could support clinical decisions to increase
the drug dose or administer a different drug
■ In silico molecular modeling can be used to predict the occurrence of all
activating but drug-resistant secondary mutations and to develop a multi-drug
targeted prevention strategy
Box 1 | History of the GIST and spectrum of its clinical presentation
The gastrointestinal stromal tumor (GIST) ‘revolution’ started in 1998 with the
demonstration that a substantial subset of lesions harbor mutations in
the gene encoding the receptor KIT leading to constitutive ligand-independent
phosphorylation.55 The second breakthrough was 5 years later, when a subset
of GISTs lacking KIT mutations harbored activating mutations in the related
PDGFRA.56,57 Approximately 5–10% of GISTs lack mutations in either kinase; in
these cases occasional BRAF mutations are reported20 and, anecdotally, KRAS
mutations. Sporadic GISTs are commonly found in the stomach (60%), small
intestine (25%), rectum (5%) and esophagus (83%).58
GISTs may develop in syndromes such as neurofibromatosis type 1 and Carney
triad or Carney/Stratakis dyad, which have wild-type KIT and PDGFRA status.
Remarkably, dyad paraganglioma and GIST occur in patients carrying germline
mutations of genes encoding succinate dehydrogenase subunits supporting the
notion that these genes, rather than constitutively active tyrosine kinases, may be
responsible for GIST formation in these patients.59 Consistently with this view, it has
been recently reported that a number of sporadic pediatric wild-type KIT/PDGFRA
GISTs carry a loss of function mutation of succinate dehydrogenase subunits.60
translocation, an autocrine–paracrine loop not mediated
by gene alteration, and signaling pathway deregulation.3
core domain of rTKs
Protein kinases transmit and amplify intracellular signals
through selective phosphorylation of residues on other
proteins, often other kinases. similar to all protein
kinases, rtKs contain a catalytic domain (core domain)
where atP binds. the binding of a specific ligand to a
docking site in the extracellular domain of the receptor,
followed by receptor dimerization via tyrosine residues
within the cytoplasmic domain, marks the first step of
receptor activation. rtK autophosphorylation4 activates
the kinase and increases its intrinsic tyrosine kinase
activity. this event creates phosphorylated tyrosine residues that become binding sites for intracellular adapter
molecules, bringing signal transduction components
close together.
the protein kinase cytoplasmic portion is structurally conserved among all serine/threonine and tyrosine
kinases, and consists of a smaller n-terminal lobe
(n-lobe) and a larger C-terminal lobe (C-lobe) connected
by a strand known as a hinge region kinase insert.5 the
n-lobe consists of a β-sheet and a conserved α-helix.
the C-lobe is mainly α-helical and contains a segment,
the activation-loop (or a-loop), which includes residues that are phosphorylated in many kinases. the core
domain that contains the atP binding site is sandwiched
between the two lobes. this domain sits beneath the
P-loop (that is, the phosphate-binding loop or glycinerich loop) that has a role in determining the shape of the
atP binding site.
in the active kinase, a characteristic DFG-motif or
catalytic triad, which is located immediately before the
a-loop, adopts a conformation whereby the aspartic acid
and phenylalanine are oriented toward the binding site
(DFG-in or open or active conformation of the kinase,
Figure 1a). Different inactive states have been identified.
one of these states is called DFG-out (closed or inactive
conformation of the kinase). under normal physiological
conditions, the intracellular portion of the kinase exists
in a regulated state with very low activity. However, in
the presence of an activating mutation, the equilibrium
between the DFG-in and DFG-out forms is shifted toward
the active open (DFG-in) conformation of the receptor.
this change allows enhanced binding of atP and, hence,
increased level of phosphorylation of the tyrosine kinase
domain.6 notably, the DFG-out conformation, where
the phenylalanine occupies part of the atP binding
site, alters the accessibility of the binding site within the
pocket (Figure 1b). the DFG-motif represents a key point
because the majority of the detected mutations destabilize the a-loop and, consequently, alter the structure of
the catalytic triad. moreover, several drugs target specifically either the DFG-in (for example, dasatinib) or the
DGF-out (imatinib, nilotinib and sunitinib) form.
a new class of tyrosine kinase inhibitors is currently
in development. these new molecules target the socalled ‘switch pocket’, a distinct region adjacent to the
atP binding site of the protein that is unique for each
tyrosine kinase (or its subfamily). this region is involved
in the regulation of the tyrosine kinase catalytic activity:
the phosphorylation of specific amino acids alters the
domain conformation inducing a switch of the structure
of the domain itself.7,8
Mutations in KiT and Pdgfra
KIT and PDGFRA are both located on chromosome 4q12
and encode receptors belonging to the type iii tyrosine
kinase family and share a similar structure. the cognate
ligands are stem-cell factor and platelet-derived growth
factor (PDGFa and B), respectively. KIT and PDGFRA
mutations are causally involved in the development of
Gists and the response to current therapies. they are
also believed to be mutually exclusive and are subdivided
into two categories: primary, linked to pathogenesis, and
secondary, related to treatment or disease progression.
about 80% of Gists harbor primary KIT mutations that
generally occur in the juxtamembrane domain (exon 11,
termed ‘mutational hot-spot’), whereas the majority of
PDGFRA mutations (about 65%) affect the tyrosine
kinase 2 domain (exon 18, Figure 2). all of the mutations that cause rtK activation affect the downstream
162 | MARCH 2011 | voluMe 8
www.nature.com/nrclinonc
© 2011 Macmillan Publishers Limited. All rights reserved
focuS on PErSonalizEd MEdicinE
a
b
1
1
2
2
ATP
DFG-motif
DFG-motif
c
1
2
Imatinib
Imatinib
DFG-in
DFG-out
DFG-in
DFG-out
Figure 1 | KIT core conformations. a | Active kinase. (1) 3D model of the open form of KIT governed by the DFG motif.
(2) Zoom on the DFG-in motif (open, active form). D810 is colored green, F811 blue, and G812 red. ATP is depicted in atomcolored sticks (C, gray; O, red; N, blue, P, orange). Hydrogen atoms are omitted. b | Inactive kinase. (1) 3D model of the
closed form of KIT. (2) Zoom on the DFG-out motif (closed, inactive form). Note that F811 (dark blue) points inwards into the
kinase ATP-binding pocket, thereby preventing ATP binding and kinase phosphorylation. The ATP is represented by its van
der Waals surface (light purple). c | In silico 3D structures of (1) wild-type and (2) mutatated-Δ559 KIT in complex with
imatinib. The juxtamembrane hairpin motif (dark blue) is highlighted in the red circles. Imatinib is in atom-colored sticks,
(C, gray; O, red; N, blue), and its molecular surface is in orange. The α-helix C (orange), P-loop (green), and A-loop (dark red)
are also highlighted. Hydrogen atoms, water molecules, ions and counterions are omitted for clarity. Arrows depict the
equilibrium of the kinase between the open and closed form.
Pi3K/aKt and ras/maPK pathways and mammalian
target of rapamycin (mtor) and its effectors (s6 kinase
and 4eBP1).
Primary activating mutation types
a wide spectrum of Gist primary mutations has been
reported. these mutations can be divided into two categories on the basis of their location: mutations of the
receptor extracellular and cytoplasmic juxtamembrane
domains, and mutations of the two enzymatic domains
(tyrosine kinase 1 and tyrosine kinase 2, Figure 2).
Primary KIT mutations have been reported to cluster
in the extracellular (exon 9), juxtamembrane (exon 11),
tyrosine kinase 1 (exon 13), and tyrosine kinase 2
(exon 17) domains. mutations in aggregate primary
exon 13 and 17, account for 1–2% of the total mutations.9,10 mutations in KIT exon 9, followed by those
in exon 17 and 13 are overrepresented among intestinal Gists and correlate with spindle-cell morphology.
Primary PDGFRA mutations are mainly identified in
exon 18, followed by exon 12 and 14. PDGFRA mutations
occur almost exclusively in Gists of the stomach and
omentum and correlate with epithelioid morphology. all
of these mutations lead to conformational changes that
ultimately perturb the 3D structure of the receptor.
imatinib (Gleevec®, novartis aG Corporation, Basel,
switzerland) is the first effective small-molecule tyrosine
kinase inhibitor originally approved for the treatment of
chronic myeloid leukemia. Besides inhibiting BCR-ABL,
imatinib blocks the activity of several other tyrosine
kinases, including Kit and PDGFra. the recognition
of the inactive conformation of Kit and PDGFra by
imatinib enables it to bind in a pocket largely coincident
with the atP-binding site, thus preventing kinase activation by restricting the conformational transition of
the a-loop.6
imatinib response depends on KIT and PDGFRA
mutational status. Gists carrying KIT exon 11 mutations (juxtamembrane domain) respond much better to
targeted treatment than tumors with exon 9 mutations
(extracellular domain) or wild-type KIT and PDGFRA. By
contrast, primary mutations affecting the tyrosine kinase
domains do not generally respond to imatinib.11 From this
evidence, it seems that KIT and PDGFRA mutational
status has a strong predictive value. moreover, among the
primary responsive mutations of the regulatory domains
(extracellular and juxtamembrane), the drug-response
modulation leads to primary imatinib resistance.
in-frame deletions including the deletion of valine 559
(Δ559) are the most frequently detected primary mutations affecting KIT exon 11.9 this mutation induces a
substantial modification in the conformation of the
juxtamembrane domain that, in turn, results in a shift of
the dynamic equilibrium of the kinase toward the open,
active form (DFG-in, Figure 1c). when the receptor is
in its closed form, the mutated juxtamembrane is able
to better accommodate imatinib, thereby enhancing
binding affinity. this finding is consistent with in silico
nature reviews | clinical oncology
volume 8 | marCH 2011 | 163
© 2011 Macmillan Publishers Limited. All rights reserved
rEViEWS
KIT
PDGFRA
Immunoglobulin-like
domains
Extracellular domain
Exon 9
Transmembrane domain
Exon 11
Exon 12
Exon 13/14
K642E
V654A
T670I/E
Exon 17/18
C809G
D816V/Y/H/G/E/A/F
D820G/Y/E/A/V/H/N
N822K/Y/H/D
Y823D
A829P
Exon 14
H687Y
Juxtamembrane domain
Tyrosine kinase I domain
Kinase insert
Exon 15/16
D716N
K786N
Tyrosine kinase 2 domain
Exon 18
D842V
Figure 2 | KIT and PDGFRA mutations and correllation to protein structure. KIT and PDGFRA mutations are subdivided into
primary mutations (red), present at the disease onset, and secondary mutations (blue), which develop during treatment.
The reported frequency of primary and secondary mutations is represented by the thickness of the rectangles. All types of
primary mutations mostly affect exon 11. Occasionally, point mutations may occur in exons 9, 13, and 17. Insertion
mutations are very rare. The only deletion reported to occur outside the KIT juxtamembrane domain in gastrointestinal
stromal tumors involves exon 14 (tyrosine kinase 1). Secondary mutations mainly affect KIT. Over 10 imatinib-resistant
secondary mutations have been reported, most of which are missense mutations. KIT secondary mutations mainly occur in
the activation loop and are likely to increase the propensity of the enzyme to adopt its active form. The observed secondary
PDGFRA mutations are D842V (equivalent to the KIT D816E/H) and H687Y, both missense mutations.
prediction. molecular modeling shows how the sterical
hindrance exerted by the hairpin portion of the juxtamembrane domain at the atP pocket entrance, which
partly interferes with imatinib binding to the wild-type
receptor, is removed in mutated exon 11 KIT. this mutation results in a higher affinity of the mutated isoform
for imatinib with respect to its wild-type receptor
(Figure 1c). an exception is represented by the l576P
mutation of KIT exon 11 affecting the juxtamembrane
domain, which shows a minimal response rate to imatinib (Figure 3a–c), an observation fully supported by
molecular modeling prediction.12 accordingly, computer
simulations of the l576P mutant KIT in complex with
imatinib reveal that the l576P mutation imposes significant energetic constraints on several amino acids of the
imatinib binding pocket. as a consequence, the atP
pocket shape is altered, and the strong binding of imatinib is no longer preserved (supplementary Figures 1
and 2 online). this evidence is substantiated by the corresponding in silico predicted free energy of binding
value (supplementary Box 2 online).
regarding the prognostic value of KIT and PDGFRA
mutational status, data based on pre-imatinib clinical
and molecular findings indicated that untreated patients
carrying KIT exon 11 deletions had a worse outcome
and a high risk of metastasis.9,13 nonetheless, point
mutations or duplications that occur in the same exon
correlate with less-aggressive tumor behavior, similar
to most PDGFRA mutations.14,15 in any case, complete
radiological responses are rare (<5%) and after a median
of about 2 years an acquired resistance to imatinib is
generally observed.
Mechanisms of secondary resistance
a number of mechanisms have a role in Gist imatinibresistance. such mechanisms include the pharmacokinetic metabolic variability, alterations in the transporter
enzymes, and the hitherto unexamined contribution of
KIT and/or PDGFRA polymorphisms or copy number
alterations,16 which have already been demonstrated in
other histotypes.17 However, several crucial resistance
mechanisms have been envisaged and reported.
First, pathological activation of downstream signaling pathways such as Pi3K/aKt18 and ras/raF/meK/
maPK can occur, both converging on mtor.19 a BraF
mutation-based activation pathway may occasionally
occur (7% of wild-type Gists) in both imatinib-naive
and imatinib-resistant Gists without KIT and PDGFRA
mutations. in Gists where the activation of BraF triggers maPK pathway, meK inhibitors could be included
in the treatment plan.20 similarly, in neurofibromatosis
type 1 Gists the activation of the maPK cascade is
achieved through the loss of neurofibromin, which is a
negative regulator of ras signaling.21
second, activation of an alternative tyrosine kinase
receptor (for example, the oncogenic rtK, aXl) and
loss of Kit expression may occur. overexpression of
aXl induces imatinib resistance through a kinase switch
from Kit to aXl. strong aXl expression and lack of
Kit expression have been demonstrated in two imatinibresistant patients with Gists.22 molecular modeling
and an in vitro assay using the novel Kit–aXl kinase
inhibitor mP470 confirmed the results showing no
imatinib binding to mutated Kit but efficient binding
to mP470.22
164 | MARCH 2011 | voluMe 8
www.nature.com/nrclinonc
© 2011 Macmillan Publishers Limited. All rights reserved
focuS on PErSonalizEd MEdicinE
a
b
Pre-imatinib
Post-imatinib
Hematoxylin
and eosin
Mib-1
KIT
c C+
pz
1
Anti-PTyr
2
Anti-KIT
dd
KIT
e
e
K623
V654
T670
I670
C673
f
f
E640
A654
Y823
D677
Figure 3 | Imaging, biochemical and molecular evidence and modeling of KIT. a–c | L576P KIT primary mutation: receptor
activation and imatinib resistance. a | CT scan of a peritoneal nodule near to the duodenum (red circle) in progression after
imatinib 400 mg/day. b | Surgical specimen lacking any evidence of imatinib response and showing evidence of high
cellularity, high rate of the proliferative marker Ki-67 (Mib 1), and dot-like KIT decoration. c | (1) Immunoprecipitation and
(2) Western blotting analysis performed on a matched pair of frozen tissue (pz) confirms KIT activation and expression.
d–f | Comparison of d | T670I and e,f | V654A secondary KIT mutations. d | The larger molecular volume of the mutated
isoleucine (I670) side chain with respect to the wild-type threonine (T670) induces imatinib (green and red sticks) to
assume a shifted position within the pocket. Residues affected by the mutation are depicted in green and red sticks and
balls. e,f | The decrease in imatinib binding affinity is mainly due to the smaller molecular volume of alanine (A654)
compared with valine (V654) that favors the loss of packing interaction in the mutated alanine variant compared with the
wild-type counterpart. Surface complementarity between imatinib (atom-colored sticks) and either e | wild-type valine (green
sticks) or f | mutated alanine (gold sticks).
third, amplification or loss of KIT and/or PDGFRA23
may occur in rare instances.24,25 this type of alteration may
also be observed at disease onset (primary resistance).
Fourth, resistance caused by dedifferentiation (that
is, the histological progression to, or change to a higher
grade sarcoma) in the absence of secondary mutations
may be observed. indeed, disease progression and resistance have been described in advanced imatinib-treated
Gists harboring primary KIT and PDGFRA mutations
in the absence of secondary resistant mutations but in the
presence of evidence of dedifferentiation.26
Fifth, under drug pressure, the acquisition of secondary
KIT or PDGFRA mutations represents the most important and frequent mechanism of secondary resistance.
remarkably, this resistance, which correlates with clinical
progression, is due to the outgrowth of multiple resistant
clones that often contain different secondary mutations.25
Structural perturbations
secondary mutations (Box 2 and Figure 2) in the kinase
domain of KIT and sometimes in PDGFRA, are accompanied by concomitant reactivation of the corresponding tyrosine kinase even in the presence of imatinib. the
structural alterations induced by these mutations may
confer drug resistance by two distinct pathways. one such
pathway introduces a perturbation in the general architecture of the atP pocket, as is the case of the two missense mutations t670i and v654a affecting the tyrosine
kinase 1 domain27 (Figure 3d–f ). the other pathway
induces a transition from the autoinhibited (closed)
form toward the activated (open) form, as exemplified
by D820n (supplementary Figure 3 online) and D816v
mutations in the tyrosine kinase 2 domain of Kit.28
the molecular mechanisms underlying these two
pathways can be analyzed to a finer level using molecular modeling. For instance, our computer studies clearly
show that, for the first pathway, the t670i mutation
induces substantial modifications in the atP-binding
pocket of the imatinib-sensitive Kit (Δ559), resulting in
a calculated free energy of binding difference (ΔΔGbind,
supplementary Box 1 onlin) for imatinib for the wild
type Kit of –3.84 kcal/mol.27 By contrast, the v654a
mutation shows only limited modifications, leading to a
ΔΔGbind for imatinib equal to –1.52 kcal/mol. Consistently,
Δ559 + t670i Kit was completely insensitive to all tested
imatinib doses, whereas Δ559 + v654a Kit was sensitive
to the highest imatinib dose tested.
in the second pathway, the most common primary
mutation affecting PDGFRA exon 18, the D842v point
mutation (homologous to D816v in KIT), is insensitive
nature reviews | clinical oncology
volume 8 | marCH 2011 | 165
© 2011 Macmillan Publishers Limited. All rights reserved
rEViEWS
Box 2 | Secondary mutations in GISTs
All secondary mutations affect KIT with only two reported exceptions where
acquired resistance was due to PDGFRA alterations in gastrointestinal stromal
tumors (GISTs) characterized by KIT mutations.24,61 These mutations occur
after a long period of imatinib treatment (median 27 months). They are typically
observed in patients with KIT-mutated exon 11 GISTs,62 are rare in KIT exon 9
and do not generally occur in GIST patients with a wild-type genotype.25,34 Other
mutations in different lesions or simultaneous evolution of multiple clones in
one lesion have been reported.63 Combining the highly sensitive, allele-specific
PCR with denaturing high-performance liquid chromatography (D-HPLC) technique
can improve the sensitivity of mutation detection from about 15% to about 5%.
KIT-tyrosine kinase inhibitor-resistant mutations were detected in 90% of tumor
samples after treatment using these techniques (against the previously reported
45–70%).25,62,64 Furthermore, in three cases, two (34%) secondary KIT mutations
were observed in the same metastasis. Notably, patients treated with sunitinib
showed a wider spectrum of mutations (1–5) compared with those treated with
imatinib (1–2), and displayed a prevalence of KIT exon 17 mutations (60%),
in keeping with in vitro studies showing that KIT activation loop mutations are
sunitinib resistant.
to imatinib because it affects the a-loop of the catalytic
domain and leads to constitutively activated kinase.6
in in silico experiments with this mutation, the mutant
receptor has ΔΔGbind of –4.43 kcal/mol and is less likely
to bind to imatinib than the wild-type protein. notably,
the imatinib insensitivity is dramatically reversed by a
deletion of the same residue in ΔDimH842–845, once
again underscoring the importance of structural state
resulting from the changes in the residues involved in
this specific case (supplementary Figure 4 online).29 the
in-frame deletion of the four residues D842–H845 in
the core domain of PDGFra does not negatively interfere with the conformation of the imatinib binding site;
on the contrary, it is beneficial to the binding in that it
favors a better accommodation of the inhibitor within the
kinase binding pocket. accordingly, the calculated affinity of this PDGFra mutant isoform for the inhibitor is
even slightly higher than that of the wild type receptor
(ΔΔGbind = +0.89 kcal/mol).
Molecular modeling in clinical settings
molecular dynamic simulations of imatinib at the atomic
level in complex with Kit receptors with t670i and
v654a mutations indicate that these missense substitutions alter the conformation of the receptor drug-binding
pocket, although to a different extent. in fact, the replacement of a threonine with an isoleucine at position 670
of Kit introduces several substantial modifications in
the conformation of other residues, which induces a collapse of the atP-pocket. Conversely, the presence of an
alanine replacing a valine at position 654 of the same
rtK results in only moderate structural alterations.27
thus, although both t670i and v654a missense mutations cause imatinib-acquired resistance, the former is
far more resistant to imatinib than the latter. on the
basis of these data, it was predicted that the effect of the
v654a mutation might be subverted by a dose escalation
of the inhibitor that was expected to recapture a clinical response27 (Figure 3d–f). this example clearly illustrates how the tight coupling of biochemical analysis of
mutated receptors (testing the actual imatinib resistance)
and molecular modeling can yield vital information to
medical oncologists, and even indicate the most suitable
dose for escaping secondary resistance.
molecular modeling provides more than just the criteria
to determine, at a molecular level, the success or failure of a
given tyrosine kinase targeted therapy in Gist. molecular
modeling may have a more important role in overcoming
mutation resistance, as highlighted in the t670X model.30
negri et al.30 assessed why only isoleucine is found in
place of threonine at position 670 of Kit in unresponsive patients with Gist, and investigated all the mutations
permitted by genetic code at this rtK position. the six
different alternatives to the wild-type residue threonine
produced six functionally different Kit receptors; these
included two inactivated Kits (t670r and t670P, lossof-function substitutions), and four constitutively activated Kits (t670a, t670s, t670K and t670i, activating
mutations). notably, all mutants in this second set were
found to be imatinib-sensitive except for the one with the
isoleucine mutant residue, which was totally imatinibresistant.30 according to this model, we can assume that a
Darwinian-type selection is occurring because, among all
possible variant forms allowed by the genetic code, only
those clones providing a selective advantage emerge in the
presence of the drug. all of these findings were confirmed
by in silico experiments and the thermodynamic information gathered from this molecular model completely match
the in vitro data.30 taken together, the evidence favors the
utility of such an in silico approach in a clinical setting for
predicting all the possible mutations, including those not
yet detected in patients with Gist.
RTK inhibitors
the essential role of the atP pocket in maintaining
kinase function is beginning to be eludicaded, making
it a major focus for kinase inhibitor research. most of
the current inhibitors inactivate Kit by binding to the
tyrosine kinase 1 domain (exons 13 and 14) and locking
the complex in the closed conformation (DFG-out).
However, because the similarities in binding sites of different kinases at the common substrate atP may limit their
selectivity, other sites of the receptor have been exploited.
a study on structural mapping of kinase genetic variants
and their mutants coupled with crystal structure data
showed that kinase cancer mutations preferentially cluster
in the P-loop and a-loop (mutational hot-spots), and
that the most relevant mutations destabilize the inactive
tyrosine kinase form favoring a stabilization of its active
form.31 in fact, mutations in the a-loop are an ongoing
pharmacological challenge. the a-loop acts like a gate
swinging back and forth, promoting the protein conformational change from the inactive to the active conformation. thus, the conformational flexibility of these proteins
promotes the emergence of resistance.
the shift of equilibrium away from inactive state is
most certainly believed to be the key determinant of the
loss in binding capacity of the inhibitor to the receptor.
in fact, mutations in the tyrosine kinase 2 domain generally contribute to maintaining the gate in its open state
(DFG-in). licensed kinase inhibitors that bind to the
166 | MARCH 2011 | voluMe 8
www.nature.com/nrclinonc
© 2011 Macmillan Publishers Limited. All rights reserved
focuS on PErSonalizEd MEdicinE
open or active form are classified as type i inhibitors,
for example, dasatinib, whereas those that bind to the
closed or inactive form are termed type ii inhibitors, for
example, imatinib, nilotinib and sunitinib.
Sunitinib
sunitinib has been proved to be effective in imatinibresistant Gists that have secondary mutations in the
atP binding domain (v654a, exon 13 and t670i,
exon 14), whereas mutations within the a-loop seemed
to be resistant (D816v and D816H on exon 17) to the
drug. 32 experimental measurements of kinase autoactivation rates together with molecular modeling seem
to support the following hypothesis: the shifting of the
a-loop towards an open, active conformation, coupled to
an accelerated autophosphorylation of the mutant rtK,
is the mechanism underlying the resistance to sunitinib
exhibited by D816H and D816v mutants.33 on the other
hand, the binding mode of sunitinib to Kit is slightly
different from that of imatinib, the former reaching
less deep into the tyrosine kinase binding pocket than
the latter. accordingly, in the presence of sunitinib, the
longer side chain of the isoleucine residue in the Kit
t670 mutant can fit in the cavity without inducing a
dramatic conformational change of the binding site and,
therefore, without impairing substantially the affinity
of the protein for the inhibitor.33 notably, the clinical
response for primary KIT exon 9 mutant and wild-type
Gists was higher in patients treated with sunitinib than
in those treated with imatinib.34 molecular modeling
studies are ongoing to shed light on these findings.
nilotinib
among the second generation Kit and PDGFra kinase
inhibitors, nilotinib (which is structurally related to imatinib), seems to be 30 times more potent than imatinib
against wild-type KIT. nilotinib has also shown encouraging results in patients who failed imatinib and sunitinib
treatment.35 our preliminary (unpublished) calculations
seem to be in line with these clinical findings. Despite the
significant chemical and structural similarities between
imatinib and nilotinib, our in silico evidence indicates that
nilotinib assumes a slightly different conformation within
the binding pocket, allowing for a tighter fit and, hence, a
better binding. However, preliminary clinical data suggest
that substantial benefits are restricted to first-line treated
patients,36 whereas in advanced Gists no statistically
significant differences were observed in progression-free
survival or overall survival between the nilotinib and
control arm (the control arm more frequently consisted
of cases treated with imatinib or sunitinib, but also cases
for which the treatment was stopped).37
dasatinib
Dasatinib is a potent inhibitor of imatinib-resistant, wildtype, and mutated Kit. this potency is possibly a consequence of its ability to inhibit the DFG-in (active) form
of the kinase, and many of the mutations are thought to
destabilize the inactive form in favor of the active one.
in fact, dasatinib targets the active conformation, which
Table 1 | Inhibitors against alternative targets for the treatment of GIST
drug name
company
Phase of development
Drugs that inhibit HSP90 resulting in the proteasomal degradation of oncogenic client
proteins
IPI-504 (retaspimycin)
Infinity Pharmaceuticals in
conjunction with MedImmune
(Astra Zeneca)
Phase III trial was suspended
due to safety concerns
STA-9090
Synta Pharmaceuticals Corp.
Phase II
BIIB021
Biogen Idec
Phase II
BIIB028
Biogen Idec
Phase I
SNX-5422
Serenex
Phase I
XL888
Exelixis
Phase I
AUY922
Novartis
Phase I
AT13387
Astex Therapeutics
Phase I
Drug that inhibits the targeted proteolysis via the 26S proteasome
Bortezomib (Velcade)
Millenium Pharmaceuticals
Currently in combination trials
Everolimus (RAD001)
Novartis
Phase II
Ridaforolimus
(Deforolimus, AP23573)
Ariad Pharmaceuticals
Phase III
Drugs that inhibit mTOR
Drug that inhibits PI3K and mTOR
PF-04691502
Pfizer Oncology
Phase I
Drugs that deacetylate by histone deacetylase inhibitors, leading to an accumulation
of both hyperacetylated histones and transcription factors
Vorinostat (Zolinza, SAHA)
Patheon, Inc. (Merck)
Phase I
CUDC101
Curis, Inc.
Phase I
Panobinostat
Novartis
Phase I
Abbreviation: GIST, gastrointestinal stromal tumor; HSP90, heat shock protein 90; mTOR, mammalian
target of rapamycin.
is highly conserved and thus shared by other kinases.38
notably, dasatinib is a potent inhibitor of D816v and
D816F39 as well as PDGFra D842v, as demonstrated
by ex vivo and in vitro experiments.40 Crystallization of
dasatinib in different tyrosine kinase complexes clearly
reveals that this drug binds to the active state of the
kinase. therefore, all effects caused by the presence of
mutations that disrupt the inactive state of the kinase (by
reverting the conformation of the a-loop toward an open
conformation) may affect the affinity of dasatinib for the
corresponding mutant isoforms.
other kinase inhibitors
the sensitivity of KIT T670I, KIT V654A and PDGFRA
D842V mutations to PKC412 was tested in cell cultures:
these mutated receptors proved to be sensitive to the
treatment.24 moreover, PKC412 was a potent inhibitor of
imatinib-resistant D816v Kit.41,42 our in silico models
of Kit and PDGFra reveal that the side chain of D816 in
Kit is involved in a hydrogen bond with n819, which has
an essential role in maintaining the correct DGF-out conformation of the inactive kinase form. Clearly, this bond
is no longer present when valine is substituted at this
position. owing to the repositioning of the side chain,
other important hydrogen-bonds (that is, with a597
and K818) are no longer dominant interactions. the
nature reviews | clinical oncology
volume 8 | marCH 2011 | 167
© 2011 Macmillan Publishers Limited. All rights reserved
rEViEWS
a
b
Molecular data from patients
resistant to selected drugs
(e.g. imatinib)
Licensed drugs
Three-dimensional model
Predicted resistance
mutations to imatinib tyrosine
kinase mutation 1/2/3
From compound library
Drug 1/2/3
Selected drugs (with favorable free-energy binding value)
Validation on cell lines or mice
Identification of the most suitable compound
(with favorable free-energy binding value)
Proposed treatment: ‘preventive drug cocktail’
Imatinib-sensitive tyrosine kinase domain
+ Drug 1/2/3
Personalized medicine
Figure 4 | Clinical prediction of treatment outcomes with molecular modeling. a | Current approaches. The type of
secondary mutation in KIT and/or PDGFRA is assessed, followed by three-dimensional modeling of the candidate receptor
carrying the mutation of interest. Licensed drugs are screened in order to choose the compound with the most favorable
free energy of binding value. b | The new strategy includes a computer-based approach in order to predict all new possible
activating but drug-resistant receptor tyrosine kinase mutations. The three-dimensional models of the receptors carrying
these mutations will be used to screen compound libraries, select and develop drugs to be combined in a preventive
cocktail capable of inhibiting all the resistant mutations.
calculated free energy of binding for Δ559 + D816Kit
with imatinib is –10.02 kcal/mol, whereas the corresponding value for Δ559 + v816Kit with imatinib is
–6.89 kcal/mol. these results clearly indicate that mutation of the key aspartic acid residue, D816Kit, is instrumental in destabilizing the inactive state of this protein,
resulting in a lower affinity for imatinib.
PKC412 is a competitive inhibitor that targets the
open kinase form. its mechanism of action is, therefore,
clearly different from that of imatinib, which targets the
kinase inactive form. in the open conformation, the 816
residue does not seem to be involved in any particular
intramolecular interaction, and the calculated free energies of binding for the wild type and mutant forms are very
similar. accordingly, the D842v mutation does not seem
to interfere with PCK412 binding to PDGFra, in keeping
with the clinical response to this drug.43
regarding the switch pocket kinase inhibitors, if they
are able to target the switch pocket kinase domain, in
principle they should also prevent the binding of the
phosphorylated switch by competitively binding to this
region and thus blocking the conformational activation
of the kinase. the compounds currently in development
seem to be highly effective, even at a very low concentration, not only against wild-type Kit and PDGFra but
also against the most common Kit and PDGFra drugresistant mutations (v654a, t670i, D816H, D816v and
D842v) detected in recurrent patients treated with the
clinically available tyrosine kinase inhibitors.44
Alternative targets
inhibitors in development for Gist treatment are listed in
table 1. in particular, in vitro data suggested that iPi-504,
a heat shock protein 90 inhibitor, might be a therapeutic
option for Gist harboring D842v.40,45 among the new
treatment options, the agents in development are histone
deacetylase inhibitors,46 and bortezomib, which inhibits
Kit through transcriptional downregulation with strong
apoptotic effects. 47 another promising compound is
raD001, which inhibits mtor and is now in phase ii
testing.48,18 However, even though this compound displays an acceptable toxicity profile, its efficacy in the clinical setting seems to be limited in patients with advanced
Gists49 and those treated with imatinib and sunitinib.50
unfortunately, in silico data does not help to elucidate
why the efficacy is limited.
Conclusions
with the exception of two recently described Kit
mutations (s709F and K818r),51,52 the most commonly
reported Kit/PDGFra secondary mutations affect 13
different positions along the protein primary sequence.
these mutations cluster into three domain regions
(tyrosine kinase 1, kinase insert, tyrosine kinase 2), the last
of which accounts for six different amino acidic substitutions in Kit: C809, D816, D820, n822, Y823, a829.9,53
these substitutions result in 21 variants. However, from
our experimental and in silico observations, we know
that, from a functional standpoint, the overwhelming
majority of these mutations exert similar effects in that
their presence induces a change of the kinase toward its
active (open) conformation, thus narrowing the number
of atP-competitive inhibitors. the challenge is, therefore,
to overcome the development of secondary mutations
that, in a tumor as heterogeneous as Gist, can occur in
different combinations of up to five different types in the
same patient (Box 2).
the promising switch pocket inhibitors currently in
development aim to target kinases directly. However, other
proposed new strategies are aimed at inducing Kit oncoprotein degradation to obtain a strong apoptotic effect 47
or inhibiting Kit-dependent Pi3K downstream signaling
using mtor inhibitors.18 nonetheless, we think that these
proteins constitute extremely promising targets. realizing
this promise is dependant on the synergistic action of
alternative and more powerful techniques, including
computer-based approaches such as those outlined here,
to promptly predict new, possible rtK mutations.
168 | MARCH 2011 | voluMe 8
www.nature.com/nrclinonc
© 2011 Macmillan Publishers Limited. All rights reserved
focuS on PErSonalizEd MEdicinE
the Darwinian selection of mutated variants of target
proteins could be harnessed with the intelligent design
of drugs directed against the single new variants. this
innovative multi-drug targeted prevention, which can
potentially change the natural history of a disease, has
been proposed in the setting of chronic myeloid leukemia. the positive results obtained with the newly
introduced drugs nilotinib and dasatinib suggested that
a “combination of two or three kinase inhibitors, when
carefully selected to cover all known resistant mutations,
could shut off all mechanisms of escape”54 (Figure 4). a
word of caution is mandatory because additional adverse
effects caused a synergism in toxic effects may require
drug concentrations to be tailored ad hoc. However, by
predicting the mutations, and classifying them according to their degree of probability, it could be possible to
administer a drug as soon as patients display such mutations, so that disease recurrence can be prevented, rather
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
Kan, Z. et al. Diverse somatic mutation patterns
and pathway alterations in human cancers.
Nature 466, 869–873 (2010).
Baselga, J. Targeting tyrosine kinases in cancer:
the second wave. Science 312, 1175–1178
(2006).
Pierotti, M. A. et al. Targeted therapies: the
rare cancer paradigm. Mol. Oncol. 4, 19–37
(2010).
Healy, E. F., Johnson, S., Hauser, C. R. &
King, P. J. Tyrosine kinase inhibition: ligand
binding and conformational change in c-Kit and
c-Abl. FASEB J. 583, 2899–2906 (2009).
Liao, J. J. Molecular recognition of protein kinase
binding pockets for design of potent and
selective kinase inhibitors. J. Med. Chem. 50,
409–424 (2007).
Bikker, J. A., Brooijmans, N., Wissner, A. &
Mansour, T. S. Kinase domain mutations in
cancer: implications for small molecule drug
design strategies. J. Med. Chem. 52,
1493–1509 (2009).
Ahn, Y. M. et al. Switch control pocket inhibitors
of p38-MAP kinase. Durable type II inhibitors
that do not require binding into the canonical
ATP hinge region. Bioorg. Med. Chem. Lett. 20,
5793–5798 (2010).
Swann, S. L. et al. Biochemical and biophysical
characterization of unique switch pocket
inhibitors of p38alpha. Bioorg. Med. Chem Lett.
20, 5787–5792 (2010).
Lasota, J. & Miettinen, M. Clinical significance of
oncogenic KIT and PDGFRA mutations in
gastrointestinal stromal tumours. Histopathology
53, 245–266 (2008).
Lasota, J. et al. Clinicopathologic profile of
gastrointestinal stromal tumors (GISTs) with
primary KIT exon 13 or exon 17 mutations: a
multicenter study on 54 cases. Mod. Pathol. 21,
476–484 (2008).
Corless, C. L., Fletcher, J. A. & Heinrich, M. C.
Biology of gastrointestinal stromal tumors.
J. Clin. Oncol. 22, 3813–3825 (2004).
Conca, E. et al. Activate and resist: L576P-KIT
in GIST. Mol. Cancer Ther. 8, 2491–2495
(2009).
Miettinen, M., Sobin, L. H. & Lasota, J.
Gastrointestinal stromal tumors of the stomach:
a clinicopathologic, immunohistochemical, and
molecular genetic study of 1,765 cases with
long-term follow-up. Am. J. Surg. Pathol. 29,
52–68 (2005).
than waiting for the mutations to emerge to select the
right compound.
whether secondary resistance might be safely
avoided or delayed by the combined administration of
tKis upfront or by a sequential administration strategy
remains an open question. nonetheless, the possibility
to predict secondary resistance mutations, and hence to
select the best active inhibitors and the optimal doses
to treat the affected patients represents an effective tool to
face this problem with a measure of optimism.
Review criteria
Data for this Review were compiled by searching the
PubMed database. The search terms used were “receptor
tyrosine kinases”, “GIST” and “molecular modeling”. Only
articles published in English before July 28 2010 were
considered. The ASCO 2010 website was also searched.
14. Keun, P. C. et al. Prognostic stratification of highrisk gastrointestinal stromal tumors in the era of
targeted therapy. Ann. Surg. 247, 1011–1018
(2008).
15. Blanke, C. D. et al. Long-term results from a
randomized phase II trial of standard- versus
higher-dose imatinib mesylate for patients with
unresectable or metastatic gastrointestinal
stromal tumors expressing KIT. J. Clin. Oncol. 26,
620–625 (2008).
16. Astolfi, A. et al. A molecular portrait of
gastrointestinal stromal tumors: an integrative
analysis of gene expression profiling and highresolution genomic copy number. Lab. Invest. 90,
1285–1294 (2010).
17. Toffoli, G. et al. Genotype-driven phase I study of
irinotecan administered in combination with
fluorouracil/leucovorin in patients with
metastatic colorectal cancer. J. Clin. Oncol. 28,
866–871 (2010).
18. Bauer, S., Duensing, A., Demetri, G. D. & Fletcher,
J. A. KIT oncogenic signaling mechanisms in
imatinib-resistant gastrointestinal stromal tumor:
PI3-kinase/AKT is a crucial survival pathway.
Oncogene 26, 7560–7568 (2007).
19. Shaw, R. J. & Cantley, L. C. Ras, PI(3)K and
mTOR signalling controls tumour cell growth.
Nature 441, 424–430 (2006).
20. Agaram, N. P. et al. Novel V600E BRAF mutations
in imatinib-naive and imatinib-resistant
gastrointestinal stromal tumors. Genes Chrom.
Cancer 47, 853–859 (2008).
21. Maertens, O. et al. Molecular pathogenesis of
multiple gastrointestinal stromal tumors in NF1
patients. Hum. Mol. Genet. 15, 1015–1023
(2006).
22. Mahadevan, D. et al. A novel tyrosine kinase
switch is a mechanism of imatinib resistance in
gastrointestinal stromal tumors. Oncogene 26,
3909–3919 (2007).
23. Miselli, F. C. et al. c-Kit/PDGFRA gene status
alterations possibly related to primary imatinib
resistance in gastrointestinal stromal tumors.
Clin. Cancer Res. 13, 2369–2377 (2007).
24. Debiec–Rychter, M. et al. Mechanisms of
resistance to imatinib mesylate in gastrointestinal
stromal tumors and activity of the PKC412
inhibitor against imatinib-resistant mutants.
Gastroenterology 128, 270–279 (2005).
25. Liegl, B. et al. Heterogeneity of kinase inhibitor
resistance mechanisms in GIST. J. Pathol. 216,
64–74 (2008).
nature reviews | clinical oncology
26. Liegl, B., Hornick, J. L., Antonescu, C. R.,
Corless, C. L. & Fletcher, C. D.
Rhabdomyosarcomatous differentiation in
gastrointestinal stromal tumors after tyrosine
kinase inhibitor therapy: a novel form of tumor
progression. Am. J. Surg. Pathol. 33, 218–226
(2009).
27. Tamborini, E. et al. Functional analyses and
molecular modeling in two c-kit mutations
responsable for imatinib secondary resistence
in GIST patients. Oncogene 25, 6140–6146
(2006).
28. Foster, R., Griffith, R., Ferrao, P. & Ashman, L.
Molecular basis of the constitutive activity and
STI571 resistance of Asp816Val mutant KIT
receptor tyrosine kinase. J. Mol. Graph. Model.
23, 139–152 (2004).
29. Dileo, P. et al. Imatinib response in two GIST
patients carrying two hitherto functionally
uncharacterized PDGFRA mutations: an imaging,
biochemical and molecular modeling study. Int.
J. Cancer 128, 983–990 (2011).
30. Negri, T. et al. T670X KIT mutations in
gastrointestinal stromal tumors: making sense
of missense. J. Natl Cancer Inst. 101, 194–204
(2009).
31. Dixit, A. et al. Sequence and structure signatures
of cancer mutation hotspots in protein kinases.
PLoS ONE 4, 7485 (2009).
32. Nishida, T. et al. Sunitinib-resistant
gastrointestinal stromal tumors harbor cismutations in the activation loop of the KIT gene.
Int. J. Clin. Oncol. 14, 143–149 (2009).
33. Gajiwala, K. S. et al. KIT kinase mutants show
unique mechanisms of drug resistance to
imatinib and sunitinib in gastrointestinal stromal
tumor patients. Proc. Natl Acad. Sci. USA 106,
1542–1547 (2009).
34. Heinrich, M. C. et al. Primary and secondary
kinase genotypes correlate with the biological
and clinical activity of sunitinib in imatinibresistant gastrointestinal stromal tumor. J. Clin.
Oncol. 26, 5352–5359 (2008).
35. Montemurro, M. et al. Nilotinib in the treatment
of advanced gastrointestinal stromal tumours
resistant to both imatinib and sunitinib. Eur.
J. Cancer 45, 2293–2297 (2009).
36. Casali, P. G. et al. Preliminary data of nilotinib in
the first-line treatment of patients with
metastatic or unresectable gastrointestinal
stromal tumors (GIST) [abstract]. J. Clin. Oncol.
28 (Suppl. 15), TPS332 (2010).
volume 8 | marCH 2011 | 169
© 2011 Macmillan Publishers Limited. All rights reserved
rEViEWS
37. Reichardt, P. et al. Phase III trial of nilotinib in
patients with advanced gastrointestinal stromal
tumor (GIST): first results from ENEST G3
[abstract]. J. Clin. Oncol. 28 (Suppl. 15), 10017
(2010).
38. Lee, S. J. & Wang, J. Y. Exploiting the promiscuity
of imatinib. J. Biol. 8, 30 (2009).
39. Schittenhelm, M. M. et al. Dasatinib (BMS354825), a dual SRC/ABL kinase inhibitor,
inhibits the kinase activity of wild-type,
juxtamembrane, and activation loop mutant KIT
isoforms associated with human malignancies.
Cancer Res. 66, 473–481 (2006).
40. Dewaele, B. et al. Activity of dasatinib, a dual
SRC/ABL kinase inhibitor, and IPI-504, a heat
shock protein 90 inhibitor, against
gastrointestinal stromal tumor-associated
PDGFRAD842V mutation. Clin. Cancer Res. 14,
5749–5758 (2008).
41. Roberts, K. G. et al. Resistance to c-KIT kinase
inhibitors conferred by V654A mutation. Mol
Cancer Ther. 6, 1159–1166 (2007).
42. Ou, W. B., Zhu, M. J., Demetri, G. D., Fletcher, C. D.
& Fletcher, J. A. Protein kinase C-theta regulates
KIT expression and proliferation in
gastrointestinal stromal tumors. Oncogene. 27,
5624–5634 (2008).
43. Gotlib, J. et al. Activity of the tyrosine kinase
inhibitor PKC412 in a patient with mast cell
leukemia with the D816V KIT mutation. Blood
106, 2865–2870 (2005).
44. Heinrich, M. C. et al. In vitro activity of novel KIT/
PDGFRA switch pocket kinase inhibitors against
mutations associated with drug-resistant GI
stromal tumors [abstract]. J. Clin. Oncol.
28 (Suppl. 15), 10007 (2010).
45. Bauer, S., Yu, L. K., Demetri, G. D. & Fletcher, J. A.
Heat shock protein 90 inhibition
in imatinib-resistant gastrointestinal stromal
tumor. Cancer Res. 66, 9153–9161 (2006).
46. Muhlenberg, T. et al. Inhibitors of deacetylases
suppress oncogenic KIT signaling, acetylate
HSP90, and induce apoptosis in gastrointestinal
stromal tumors. Cancer Res. 69, 6941–6950
(2009).
47. Bauer, S. et al. Proapoptotic activity of
bortezomib in gastrointestinal stromal tumor
cells. Cancer Res. 70, 150–159 (2010).
48. Schoffski, P. et al. A phase I-II study of everolimus
(RAD001) in combination with imatinib in
patients with imatinib-resistant gastrointestinal
stromal tumors. Ann. Oncol. 21, 1990–1998
(2010).
49. Hohenberger, P. et al., Multicenter, single-arm, twostage phase II trial of everolimus (RAD001) with
imatinib-resistant patients (pts) with advanced
GIST [abstract]. J. Clin. Oncol. 28 (Suppl. 15),
10048 (2010).
50. Richter, S. et al., Multicenter, triple arm, singlestage, phase II trial to determine the efficacy and
safety of everolimus (RAD001) in patients with
refractory bone or soft tissue sarcomas including
GIST [abstract]. J. Clin. Oncol. 28 (Suppl. 15),
10038 (2010).
51. Wardelmann, E. et al. Polyclonal evolution of
multiple secondary KIT mutations in
gastrointestinal stromal tumors under treatment
with imatinib mesylate. Clin. Cancer Res. 12,
1743–1749 (2006).
52. Grabellus, F. et al. Double resistance to imatinib
and AMG 706 caused by multiple acquired KIT
exon 17 mutations in a gastrointestinal stromal
tumour. Gut. 56, 1025–1026 (2007).
53. Nishida, T. et al. Secondary mutations in the
kinase domain of the KIT gene are predominant
in imatinib-resistant gastrointestinal stromal
tumor. Cancer Sci. 99, 799–804 (2008).
54. Sawyers, C. L. Even better kinase inhibitors for
chronic myeloid leukemia. N. Engl. J. Med. 362,
2314–2315 (2010).
55. Hirota, S. et al. Gain-of-function mutations of c-kit
in human gastrointestinal stromal tumors.
Science 279, 577–580 (1998).
56. Heinrich, M. C. et al. PDGFRA activating
mutations in gastrointestinal stromal tumors.
Science 299, 708–710 (2003).
57. Hirota, S. et al. Gain-of-function mutations of
platelet-derived growth factor receptor alpha
gene in gastrointestinal stromal tumors.
Gastroenterology 125, 660–667 (2003).
170 | MARCH 2011 | voluMe 8
58. DeMatteo, R. P. et al. Two hundred
gastrointestinal stromal tumors: recurrence
patterns and prognostic factors for survival.
Ann. Surg. 231, 51–58 (2000).
59. Pasini, B. et al. Clinical and molecular genetics of
patients with the Carney-Stratakis syndrome and
germline mutations of the genes coding for the
succinate dehydrogenase subunits SDHB, SDHC,
and SDHD. Eur. J. Hum. Genet. 16, 79–88 (2008).
60. Janeway, K. A. et al. Succinate dehydrogenase in
KIT/PDGFRA wild-type gastrointestinal stromal
tumors [abstract]. J. Clin. Oncol. 28 (Suppl. 15),
10008 (2010).
61. Lim, K. H. et al. Molecular analysis of secondary
kinase mutations in imatinib-resistant
gastrointestinal stromal tumors. Med. Oncol. 25,
207–213 (2008).
62. Antonescu, C. R. et al. Acquired resistance to
imatinib in gastrointestinal stromal tumor occurs
through secondary gene mutation. Clin. Cancer
Res. 11, 4182–4190 (2005).
63. Heinrich, M. C. et al. Molecular correlates of
imatinib resistance in gastrointestinal stromal
tumors. J. Clin. Oncol. 24, 4764–4774 (2006).
64. Desai, J. et al. Clonal evolution of resistance to
imatinib in patients with metastatic
gastrointestinal stromal tumors. Clin. Cancer Res.
13, 5398–5405 (2007).
acknowledgments
We wish to thank all the INT clinical staff who made
this multidisciplinary investigation of GIST possible,
in particular Dr Paolo G. Casali and Dr Alessandro
Gronchi. A special thanks goes to Dr Elena Fumagalli
who answered to all our clinical questions. The
authors are partially funded by the Associazione
Italiana per la Ricerca sul Cancro (AIRC).
author contributions
M. A. Pierotti, E. Tamborini, T. Negri, S. Pricl and
S. Pilotti contributed equally to the literature review,
discussions of the content, writing the article and to
review and/or editing of the manuscript before
submission.
Supplementary information is linked to the online
version of the paper at www.nature.com/nrclinonc
www.nature.com/nrclinonc
© 2011 Macmillan Publishers Limited. All rights reserved